Research map / Vision-language and multimodal / Multimodal LLMs
Beyond Sequential Distance: Inter-Modal Distance Invariant Position Encoding
Lin Chen, Bolin Ni, Qi Yang, Zili Wang, Kun Ding, Ying Wang, Houwen Peng, SHIMING XIANG
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add DIPE (anchored inter-modal position encoding) | MRoPE, Qwen2.5-3B · Long-context VQA (8K distractors), all 19 benchmarks avg | average accuracy: 44.69 -> 48.79 | Improved |
| add DIPE | Vanilla RoPE, Qwen2.5-3B · Long-context VQA (8K distractors), all 19 benchmarks avg | average accuracy: 46.31 -> 48.31 | Improved |
| add DIPE | MRoPE-I, Qwen2.5-3B · Long-context VQA (8K distractors), all 19 benchmarks avg | average accuracy: 46.50 -> 48.51 | Improved |
| add DIPE | MRoPE, Qwen2.5-3B · POPE (long-context 8K) | accuracy: 74.50 -> 85.58 | Improved |
| add DIPE | MRoPE-I, Qwen2.5-3B · V* (long-context 8K) | accuracy: 50.79 -> 47.64 | Hurt |
| add DIPE | MRoPE, Qwen2.5-3B · Short-context VQA, all 19 benchmarks avg | average accuracy: 51.15 -> 51.38 | No clear effect |
| add DIPE | MRoPE-I, Qwen2.5-3B · Short-context VQA, all 19 benchmarks avg | average accuracy: 51.49 -> 51.41 | No clear effect |
| add DIPE | MRoPE, Qwen3-1.7B · Long-context VQA (8K distractors), all benchmarks avg | average accuracy: 40.21 -> 44.46 | Improved |
| add DIPE | MRoPE, Qwen2.5-0.5B · Long-context VQA (8K distractors), all benchmarks avg | average accuracy: 29.51 -> 38.32 | Improved |
| add DIPE | MRoPE, Qwen2.5-3B · Interleaved image-text test (1K samples) | accuracy: 60.1 -> 60.4 | No clear effect |
| add DIPE | MRoPE, Qwen2.5-3B · Latency, attention block forward at 32K | latency (ms): 13.56 -> 14.31 | Hurt |
| add DIPE | MRoPE, Qwen2.5-3B · Latency, attention block backward at 1K | latency (ms): 0.83 -> 1.52 | Hurt |
Closest papers by meaning
- Revisiting Multimodal Positional Encoding in Vision–Language Models (ICLR 2026 · similarity 0.89)
- HoPE: Hybrid of Position Embedding for Long Context Vision-Language Models (NeurIPS 2025 · similarity 0.88)
- Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs (ICML 2026 · similarity 0.88)
- Towards Long-window Anchoring in Vision-Language Model Distillation (AAAI 2026 · similarity 0.88)
- Multi-modal Instruction Tuned LLMs with Fine-grained Visual Perception (CVPR 2024 · similarity 0.86)
- Mitigating Object Hallucination via Concentric Causal Attention (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
